[finding] turbo.json's test:e2e task is held by zero of 78 packages — the real Playwright script is test:smoke, which turbo.json does not configure at all

Measured while building check:turbo-task-graph for #12046, on 1f6b8bb193. Filed unassigned; recording the gap, not claiming it. Nothing is red today.

What was measured

turbo.json declares a generic test:e2e task:

"test:e2e": {
  "dependsOn": ["build"],
  "cache": false,
  "outputs": ["playwright-report/**", "test-results/**"],
  "inputs": ["src/**", "e2e/**", "playwright.config.ts"]
}

Enumerating every workspace package through scripts/workspace-enumerator.mjs and counting which ones declare each generic task as a script:

generic "build":          72 holder(s)
generic "test":           72 holder(s)
generic "typecheck":      65 holder(s)
generic "test:e2e":        0 holder(s)      <-- 
generic "dev":            24 holder(s)
generic "clean":           5 holder(s)
generic "gen:schema":      1 holder(s)  -> @objectstack/spec
generic "gen:skill-refs":  1 holder(s)  -> @objectstack/spec

Root package.json carries "test:e2e": "turbo run test:e2e", so the command exists, runs, matches nothing, and exits 0 — the same silent shape check:pnpm-filter-targets exists for one level over.

The repo does have a Playwright suite. examples/app-showcase holds playwright.config.ts and declares it as:

"test:smoke": "playwright test --config=playwright.config.ts"

So the task turbo.json configures cannot run, and the task that does run is unconfigured — no outputs, no inputs, no dependsOn: ["build"], which is exactly what a Playwright run needs.

Why this is not merely tidy-up

Two live consequences, neither of which reports anything today:

pnpm test:e2e at the root is a no-op that exits 0. Anyone (agent or human) who runs it as evidence that the e2e suite passes is reading a green from a command that ran nothing — check:react-declaration-parity 是唯一没接进任何 workflow 的源码审计门禁,且无 MANIFEST 时静默 skip 退出 0 —— 它现在永远不可能红 #4690's family.
pnpm turbo run test:smoke gets the default task treatment: no declared inputs, so its cache key is whatever $TURBO_DEFAULT$ happens to be, and no dependsOn: ["build"], so nothing guarantees a built app underneath it.

Why it is filed rather than fixed in place

The remedy is an edit to turbo.json, which was not in #12046's declared file surface. It is also the one row standing between check:turbo-task-graph and a wider population: that gate judges <pkg>#<task> keys only, and its header states why the generic limb is not enforced yet — the invariant is the same shape ("a task nothing can run is inert") and this is its single live violation.

Suggested disposition (not a decision)

Two readings, and they differ in what they assert about intent:

The task name is wrong. Rename test:e2e → test:smoke in turbo.json (and the root script), so the entry configures the suite that exists. inputs would want revisiting: e2e/** names a directory app-showcase does not have.
The script name is wrong. Rename app-showcase's test:smoke → test:e2e. Cheaper in turbo.json, but test:smoke is the name the checklist/dogfood tooling knows; check its callers before picking this one.

Whichever lands, widen check:turbo-task-graph's population to generic keys in the same change — the gate is written for it and its header names this issue as the precondition.

Refs: #12046 (the coverage card this was measured under) · PR for check:turbo-task-graph

Generated by Claude Code

Claim: devx@objectstack seat (#6023) · session session_01UjM2ia8Av1v5NqfqQEQmC6 · R10
Branch claude/issue-12373-turbo-e2e-task · dedicated worktree · model tier: standard.

Declared file surface: turbo.json and scripts/check-turbo-task-graph.mjs (+ its --self-test).

Serial: none. ⭐ That gate landed this shift (PR #12374, ff56a8587c) and its header names this card by number as the reason generic keys are out of population — so you are closing a residue the gate's author deliberately left open.

Symptom verified before dispatch, and a near-miss worth recording:

"test:e2e"  defined in turbo.json                              → 1
            declared by a WORKSPACE MEMBER                     → 0
            declared by the ROOT package.json                  → 1   ← not a member
"test:smoke" declared by examples/app-showcase/package.json    → 1   ← a real member

⚠️ My first count returned 1 for test:e2e and I nearly reported the card's "zero of 78" as wrong. It is not — the single holder is the root manifest, which is not one of the 78 workspace members. I counted files; the card counts members. Recording it because it is the same shape as this seat's error 27: a number measuring a related-but-different population, which reads as a refutation.

Zone 1 — rulings (scope, not facts; not renegotiable)

The surface above is the whole surface.
⭐ The gate's header states the sequencing and it is binding: widen the population to generic keys in the same change that fixes the entry, not before — "a gate that ships red is worse than no gate." So this card is both halves or neither.
⛔ Do not "fix" the entry by deleting it if the intent was real. Establish what test:e2e was meant to configure before choosing between rename, repoint, and removal.
Falsification escape: if a Zone 2 item measures false, stop and return premise_still_valid: false.

Zone 2 — assumptions; falsify before building on them

a. I assume test:smoke in examples/app-showcase is the intended target and the entry is simply misnamed. ⚠️ Not established. The outputs: ["playwright-report/**", …] suggests Playwright, and that package holds the Playwright script — but "suggests" is not "measured". Read what the script actually runs.
b. ⚠️ I assume no CI workflow invokes turbo run test:e2e today. If one does, this entry is inert and something is silently running nothing — a bigger finding. Check the workflows before editing.
c. I assume widening the gate to generic keys is safe once the entry is fixed. Measure it: a generic key held by zero members may be legitimate for keys that exist only as a dependsOn target.

Zone 3 — suggested route (not binding)

Establish intent from the script and the workflows → fix the entry → widen the gate in the same change → pin both: the fixed entry passes, and a generic key held by no member reds.

Ablate both ways, confirm on disk, restore under trap … EXIT INT TERM, prove byte-identical with git hash-object. Gate union at the final commit; re-derive on STALE TREE. Exit codes before any pipe.

Generated by Claude Code

Claim amendment: devx@objectstack seat · session session_01UjM2ia8Av1v5NqfqQEQmC6 · branch claude/issue-12373-turbo-e2e-task

File surface amended — one file added to the declared surface: the root package.json script row.

Declared surface was turbo.json + scripts/check-turbo-task-graph.mjs. The triage comment above names the root package.json script row as a landing file; the dispatch ruling's Zone 1 surface omitted it. Measured before deciding, on turbo 2.10.10, after the rename lands in turbo.json:

$ pnpm exec turbo run test:e2e --dry=json
EXIT=1
  x Missing tasks in project
  `->   x Could not find task `test:e2e` in project

So the card cannot be closed without it. Leaving the root row as "test:e2e": "turbo run test:e2e" does not preserve the status quo — it converts the silent exit-0 no-op this card is filed against into a hard exit-1 error on a command contributors run. The row moves with the task key or it breaks.

Bounded in-place, against the four conditions:

Same defect class — it is live consequence 1 of this very card ("pnpm test:e2e at the root is a no-op that exits 0").
Mechanical, correct form already pinned — the rename target is fixed by the script examples/app-showcase declares, and the triage comment already names this file as a landing file.
One line, no other claim — the single row changes; pnpm check:pnpm-filter-targets (the family that reads root scripts) is green on the result.
No new verification surface — re-derived with node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack after the edit: 3 paths, 10 matched families, no family that the two-file surface did not already carry.

Named with its evidence in the PR body, per the exemption's second obligation.

Generated by Claude Code

os-dev-report

{
  "issue": 12373,
  "status": "done",
  "branch": "claude/issue-12373-turbo-e2e-task",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/12468",
  "premise_still_valid": true,
  "summary": "Both halves landed in one change. Half 1: turbo.json's inert `test:e2e` task key is renamed to `test:smoke`, the script examples/app-showcase actually declares, and the root package.json row moves with it. Intent was established from git before choosing, per Zone 1 ruling 3: `test:e2e` entered turbo.json on 2026-05-21 (7972e7b829) when examples/app-crm declared `\"test:e2e\": \"playwright test\"` beside an e2e/ dir and a playwright.config.ts, and e737fbce39 ('simplify app-crm to minimal metadata smoke-test', 2026-05-24) deleted that script, leaving the turbo entry behind — so it is residue of a deleted script whose intent (configure the repo's Playwright suite) is still live, not a mistaken entry to delete. The rename went to the task key rather than the script because test:smoke has four live callers (showcase-smoke.yml, two platform-checklist area files, a vitest.config.ts comment) and test:e2e has zero. Half 2: check-turbo-task-graph's verdict() now judges generic (#-less) keys — a generic key must be declared by >=1 workspace MEMBER, with the root-manifest near miss written into the failure text and pinned as text in --self-test, plus a second non-vacuity floor for an empty generic population. Two card claims were falsified and are reported rather than papered over: (a) `e2e/**` does NOT name a directory app-showcase lacks — examples/app-showcase/e2e/ is tracked on origin/main with four files and playwright.config.ts declares testDir: './e2e', so all three input globs resolve (108 files hashed in the dry run) and are left as they stand; (b) Zone 2b measured SAFE — no CI workflow invokes `turbo run test:e2e`, so nothing was silently running nothing in CI (showcase-smoke.yml calls `pnpm test:smoke` with working-directory examples/app-showcase, bypassing turbo entirely — filed as #12466). Zone 2c measured: the only dependsOn targets in the whole file are `build` and `^build`, and `build` is held by 72/78, so NO generic key exists only as a dependsOn target; no exemption was added, and the header states why one would be a hole. One glob dropped inside the declared surface: `playwright-report/**`, which showcase-smoke.yml's own comment measures as never created under this config's stdout-only reporters (receipt: run 30796529117). FILE SURFACE AMENDED — the root package.json script row was added under the bounded in-place exemption, declared in a claim amendment comment on this card with its four conditions; forced by measurement, since after the rename `turbo run test:e2e` exits 1 ('Could not find task test:e2e in project'), so leaving the root row would convert this card's silent exit-0 no-op into a hard error.",
  "tests": "All results at final commit c246e4fcf6 on a clean tree (git rev-parse --short HEAD), quoted from each gate's own printed verdict line, exit codes captured before any pipe. TURBO BEHAVIOUR, measured on turbo 2.10.10 via `--dry=json`: pre-fix `turbo run test:e2e` -> EXIT=0, 78 tasks placed, 0 with a real command (the card's silent green, confirmed at tool level); post-fix `turbo run test:smoke` -> exactly 1 real command, `@objectstack/example-showcase#test:smoke` = `playwright test --config=playwright.config.ts`, outputs ['test-results/**'], cache local:false remote:false, deps ['@objectstack/example-showcase#build'], 108 input files including e2e/showcase-smoke.spec.ts; post-fix `turbo run test:e2e` -> EXIT=1 'Could not find task test:e2e in project'. REVERSE VERIFICATION, both legs, mutation confirmed on disk by grepping the exact text meant to change ('\"test:e2e\"'=1, '\"test:smoke\"'=0) and restored under trap ... EXIT INT TERM: the WIDENED gate on the pre-fix turbo.json -> EXIT=1, 'FAIL: turbo.json carries 1 inert task entry. - \"test:e2e\" is a generic task that NO workspace package declares a script for'; the PRE-WIDENING gate (origin/main's copy) on that same pre-fix tree -> EXIT=0, 'OK: 21 package-scoped turbo task(s) judged...' — which is why the residue survived three months. Restore proved by hash, not by an empty diff: git hash-object turbo.json = 3ee605711c6a62ed48bf8bdbb6c44276d12b937a = git rev-parse HEAD:turbo.json. ABLATION of the new limb: `holdersOf(key, scriptsByPackage).length === 0` mutated to `.length < 0`, confirmed on disk by counting both the removed and injected text ('=== 0': 1 -> 0; '< 0': 0 -> 1), then --self-test re-run -> EXIT=1, 4 named cases failed ('a generic task no package declares is a finding', 'the inert generic finding rules out the root manifest by name', 'a misspelled generic key is a finding', 'a misspelled generic key suggests the script that exists'); restored byte-identical (git hash-object = git rev-parse HEAD:), self-test green again. NO REBUILD LEG APPLIES: this gate is plain Node reading the repo from source (no package export/dist resolution, no vitest alias), so there is no stale-dist direction for either ablation leg — stated rather than skipped. GATE UNION re-derived at the final commit with `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack` (no hand-written path list; the script took the change set from git: 3 paths vs merge base 68c5dbaab; stderr confirmed 'objectstack-ai/objectstack at commit c246e4fcf6'), all EXIT=0: check:turbo-task-graph --self-test ('OK: all cases passed'); check:turbo-task-graph ('OK: 21 package-scoped and 8 generic turbo task(s) judged against 78 workspace package(s)'); check:nul-bytes + --self-test (6892 text files, 75 assertions); check:pnpm-filter-targets (140/177 --filter occurrences across 30 files resolve); check:cross-package-test-inputs (18 packages read outside themselves, all declared); check:ci-filter-parity (105 declared cross-package globs covered); check:agent-test-spelling; check:bash32-floor (98 cases); check:cli-command-ids (280 literals across 100 files); check:entry-guard (170 scripts/ files); check:parse-guard (46 cases). Both convention-triggered obligations for editing a gate script, run and green: `node scripts/pm/bare-root-worklist.mjs --self-test` EXIT=0 ('46 live row(s), 39 recorded verdict(s) — none stale, none missing') and `check:pm-dispatch-gates` EXIT=0 (703 cases); ROOT_FILE_WATCH_HINTS is unchanged so no new bare-root verdict is incurred. LINT — a DECLARED NARROWING with all three measurements, not a skipped step: (1) population read from eslint's own config — it reports package.json and turbo.json as 'File ignored because no matching configuration was supplied'; (2) count from --format json — 3 files requested, 1 actually linted (scripts/check-turbo-task-graph.mjs), 0 errors 0 warnings; (3) invariance — eslint.config.mjs lines 326-335 record with a positive control that this repo 'never enables type-aware linting ... for ANY file', so this diff cannot move any untouched file's verdict. NO pnpm test / pnpm typecheck is owed: no package source changed (turbo.json, one root script row, one dependency-free script in scripts/). CI is not waited on per the 2026-08-10 ruling; PR check status at report time is in_progress.",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as #12465: check:turbo-task-graph reports turbo's legitimate `//#lint` root-task spelling as an unknown package `//`, and its failure text asserts 'the override never reaches the task graph' — contradicted by a dry run showing the task in the graph with a real root command at exit 0; sits in the PRE-EXISTING package-scoped arm, untouched by this PR; filed as a sub-issue of #12046",
    "filed as #12466: the showcase smoke's only CI entry point bypasses turbo (showcase-smoke.yml uses working-directory + `pnpm test:smoke`), so the turbo.json task config repaired by this card has zero CI consumers today; observation, nothing red"
  ]
}

Generated by Claude Code

fix(devx): judge turbo's `//#<task>` root spelling against the root m…

Navigation Menu

Repository navigation

[finding] turbo.json's `test:e2e` task is held by zero of 78 packages — the real Playwright script is `test:smoke`, which turbo.json does not configure at all #12373

What was measured

Why this is not merely tidy-up

Why it is filed rather than fixed in place

Suggested disposition (not a decision)

Zone 1 — rulings (scope, not facts; not renegotiable)

Zone 2 — assumptions; falsify before building on them

Zone 3 — suggested route (not binding)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[finding] turbo.json's test:e2e task is held by zero of 78 packages — the real Playwright script is test:smoke, which turbo.json does not configure at all #12373

Description

What was measured

Why this is not merely tidy-up

Why it is filed rather than fixed in place

Suggested disposition (not a decision)

Activity

yinlianghui commented on Aug 26, 2026

Zone 1 — rulings (scope, not facts; not renegotiable)

Zone 2 — assumptions; falsify before building on them

Zone 3 — suggested route (not binding)

yinlianghui commented on Aug 26, 2026

yinlianghui commented on Aug 26, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

[finding] turbo.json's `test:e2e` task is held by zero of 78 packages — the real Playwright script is `test:smoke`, which turbo.json does not configure at all #12373